info:novinky_z_oi

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
07-03-2025:

Nové předměty magisterské specializace AI (díky @profojak za zjištění):

==== Foundation of machine learning ====
  * Empirical Risk Minimization
  * Generative learning (MLE)
  * Bayesian learning
  * Generalization theory (PAC learning)
  * Linear models (Perceptron algorithms, SVM,..)
  * EM Algorithm
  * Bias-Variance trade-off
  * Model selection and validation methods
  * Performance metrics
  * VC dimension
  * Deep learning and generalization

==== Deep learning ====
  * Recap of Machine Learning
  * Backpropagation (+implicit differentiation)
  * Convolutional Neural Networks (+equivariances)
  * Training Deep Models (initialization, normalization, residual)
  * Regularization Methods for NNs
  * Stochastic Gradient Descent (SGD)
  * Adversarial patterns. Robust learning approaches
  * Adaptive SGD methods
  * Learning Representations I: Word Vectors, Metric Learning
  * Learning Representations II: Unsupervised Representation Learning, VAE
  * Graph Neural Networks
  * Self-Attention, Transformers, ViT, LLMs
  * TBA (Recurrent NNs + State Space Models / GAN + Diffusion / Fairness)

==== Reinforcement learning ====
  * Motivation (successes, AGI, human feedback, history)
  * Multi-armed bandit problems (stochastic, contextual)
  * Solving MDPs 1: (Bellman equations, Value iteration)
  * Solving MDPs 2: (Contraction, Policy iteration)
  * Temporal difference learning 1: (TD(0), Sarsa, Q-learning)
  * Temporal difference learning 2: (n-step,Double-Q, DQN)
  * Policy gradient methods 1: (Tabular)
  * Policy gradient methods 2: (Variance reduction, Neural)
  * Combining learning and planning (AlphaZero, muZero)
  * Exploration in RL
  * Multi-agent RL (cooperative vs. adversarial)
  * Applications: Advertising, RLHF, Robotics, …
  * Neuro-science and RL

==== Computational game theory ====
  * Normal-form games
  * Two-player zero-sum games
  * Extensive-form games with imperfect information
  * Solving extensive-form games with imperfect information
  * Alternatives to Nash equilibrium
  * Online learning, regret minimization
  * Bayesian games
  * Analysis of auctions I (second/first-price auctions as instances of Bayesian games and their equilibria)
  * Analysis of auctions II (order statistics, revenue equivalence theorem)
  * Introduction to mechanism design (the framework based on Bayesian games, optimal auctions, VCG, revelation principle)
  * Coalitional games
  * Shapley value
  * Weighted voting games
  * Summary.

==== Logical reasoning and planning ====
  * Introduction and motivation + propositional logic/CNF/BDDs
  * SAT - basic encodings, DPLL, resolution
  * SAT - CDCL
  * SAT - encodings, SMT
  * SMT - lazy, congruence closure, theory combination
  * Proof assistants
  * Intro planning, heuristic search GBFS, A*, Weighted A*
  * State space representations PDDL, STRIPS, FDR
  * Delete relaxation heuristics hmax, hadd, hff
  * Landmarks, saturated cost partitioning, LM-Cut / BDD based symbolic search 

==== 6. Povinně volitelný z následující sady předmětů ====
LLMs and AGIs, Deep generative models, AI in robotics, Methods of computer vision, Non-smooth non-convex optimization for training deep neural networks
Trace: • b4b01jag • b4b38psia
Search

Navigation

Print/export

Tools

QR Code
